QianLingjun
QianLingjun
hi ,are you address this issue?Thinks
wget https://github.com/ninja-build/ninja/releases/download/v1.8.2/ninja-linux.zip sudo unzip ninja-linux.zip -d /usr/local/bin/ sudo update-alternatives --install /usr/bin/ninja ninja /usr/local/bin/ninja 1 --force
sir ,are you address this problem? i find this problem too.Thinks
hi ,do you solve this issue?Thinks
> 这样修改是可以正常训练和测试的。但是训练和测试阶段上采样的方式不一样，相当于改变了模型的结构，这样对测试结果，应该会有影响吧。
> 是最后一层的输出，应该不会有影响的，相当于resize吧
网络输入与输出的不匹配是在模型地方就存在了bug，输入256*256，但是输出最后一层是512*512，希望作者可以看下这里的问题，谢谢
> What's wrong with the model, and how do I correct the model? thinks
the reason is rpn_proposals is none, you should do tab in eval model, example: ``` if self.use_rcnn: if len(self.rpn_proposals) > 0: rcnn_crops = self.rcnn_crop(feat_4, inputs, self.rpn_proposals) self.rcnn_logits, self.rcnn_deltas = data_parallel(self.rcnn_head,...
hi, do you get a good method to address this problem?